Transcriptional profiling and network analysis of the murine angiotensin II-induced abdominal aortic aneurysm

Abstract

We sought to characterize temporal gene expression changes in the murine angiotensin II (ANG II)-ApoE−/− model of abdominal aortic aneurysm (AAA). Aortic ultrasound measurements were obtained over the 28-day time-course. Harvested suprarenal aortic segments were evaluated with whole genome expression profiling at 7, 14, and 28 days using the Agilent Whole Mouse Genome microarray platform and Statistical Analysis of Microarrays at a false discovery rate of <1%. A group of angiotensin-treated mice experienced contained rupture (CR) within 7 days and were analyzed separately. Progressive aortic dilatation occurred throughout the treatment period. However, the numerous early expression differences between ANG II-treated and control were not sustained over time. Ontologic analysis revealed widespread upregulation of inflammatory, immune, and matrix remodeling genes with ANG II treatment, among other pathways such as apoptosis, cell cycling, angiogenesis, and p53 signaling. CR aneurysms displayed significant decreases in TGF-β/BMP-pathway signaling, MAPK signaling, and ErbB signaling genes vs. non-CR/ANG II-treated samples. We also performed literature-based network analysis, extracting numerous highly interconnected genes associated with aneurysm development such as Spp1, Myd88, Adam17 and Lox. 1) ANG II treatment induces extensive early differential expression changes involving abundant signaling pathways in the suprarenal abdominal aorta, particularly wide-ranging increases in inflammatory genes with aneurysm development. 2) These gene expression changes appear to dissipate with time despite continued growth, suggesting that early changes in gene expression influence disease progression in this AAA model, and that the aortic tissue adapts to prolonged ANG II infusion. 3) Network analysis identified nexus genes that may constitute aneurysm biomarkers or therapeutic targets.

abdominal aortic aneurysm (AAA) in humans usually manifests as pathological dilation of the infrarenal abdominal aorta, often occurring on a background of atherosclerosis (ATH), thrombosis, and inflammation (7, 9). While studies have documented aortic wall gene expression differences related to these processes, the pathogenesis of AAA is incompletely understood (23, 29). Human investigations have focused primarily on gene expression in specimens from patients at “end stage,” typically when AAA diameter exceeds accepted surgical-intervention thresholds (13). To date, no reliable methods are available to stratify small aneurysms based on risk for clinical progression.

AAA animal models facilitate the study of the progressive pathophysiological changes underlying disease initiation and progression. In one such model, angiotensin II (ANG II) is infused into apolipoprotein E-deficient (ApoE−/−) mice, producing suprarenal AAAs over a 28-day time course and exhibiting predictable resultant histopathology (4, 5, 33). A proportion of mice experience contained rupture (CR) with intramural hematoma formation, most often within 10 days.

Microarray-based profiling provides simultaneous expression for thousands of genes, permitting large-scale biological pathway analysis, elucidation of new roles for known genes, and characterization of unknown genes. These methods have been used to study many aspects of cardiovascular disease (12), and chronological expression patterns can be correlated with disease progression (36).

We performed microarray analysis through a 28-day course, systematically characterizing gene expression in the ANG II/ApoE−/− model. Identifying these gene expression changes enhances our understanding of dynamic aneurysm pathology.

MATERIALS AND METHODS

Animal model.

A total of 69 male ApoE−/− mice (C57BL/6J background), 12–14 wk old, were obtained from Jackson Laboratories (Bar Harbor, ME). Alzet osmotic minipumps (model 2004; ALZA Scientific Products, Mountain View, CA) containing either angiotensin II (1,000 ng·kg⁻¹·min⁻¹, Sigma Chemical) or saline vehicle were implanted subcutaneously into mice anesthetized with isoflurane (5). All procedures were performed in accordance with Stanford University Animal Care Guidelines, and protocols were approved by the Institutional Review Board. Water and normal diet were provided ad libitum. The mortality rate observed during the study in ANG II-treated animals from presumed aortic rupture was 24%. No operative mortality occurred.

Ultrasound.

A Vevo 770 ultrasound imaging system (Visualsonics, Toronto, Canada) was used to measure in vivo abdominal aorta diameters throughout the time course using techniques similar to those described previously (38). Two-dimensional imaging at 40 MHz (B mode) was performed using a real-time microvisualization scanhead (RMV 704). Short- and long-axis scans of aortas were recorded from above the iliac bifurcation to the suprarenal region 7, 14, and 28 days after the initiation of infusion. Cine loops of 300 frames were acquired and used to determine the maximal suprarenal diameter. Two operators independently acquired and analyzed images to calculate coefficient of variation. Sizes were confirmed ex vivo after sacrificing the animals. Numbers shown represent maximal cross-sectional luminal diameters. In mice with CR, the lumen diameter included the false lumen if it was present and contiguous.

Aortic tissue collection.

After acquiring image data, we exsanguinated mouse cohorts at 7, 14, and 28 days by saline perfusion at ∼100 mmHg. Suprarenal abdominal aorta segments (consisting of the aneurysmal portion, or the equivalent region in saline controls) were surgically isolated, stripped of surrounding connective tissue, and flash-frozen in liquid nitrogen. Six mouse aortas were obtained per time point/treatment. Three additional mice with aortic CR at 7 days of ANG II, defined as mural disruption-presence of intramural hematoma on ultrasound were also harvested. Suspecting that these aortas might show different gene expression than ones without major mural disruption, we treated CR samples as their own cohort. Separate samples were derived for histological evaluation.

RNA preparation and microarray hybridization.

Aortic tissue was isolated using the RNeasy Fibrous Tissue Mini kit (Qiagen, Valencia, CA), and RNA integrity (RIN) verified by Agilent 2100 Bioanalyzer (all RIN > 7.0; Agilent Technologies, Santa Clara, CA). Unpooled total RNA (500 ng) was labeled using Cyanine-5-CTP in a T-7 transcription reaction via Agilent's Low Input Linear RNA Amplification/Labeling System as previously described (36). Labeled cRNA from samples were hybridized to the 44k Agilent Whole Mouse Genome Oligo microarray G4122F platform with an equimolar concentration of Cyanine-3-CTP-labeled universal mouse reference (Stratagene, La Jolla, CA) using standard protocols, with one mouse aorta per array. Images were quantified and feature extracted using Agilent Feature Extraction Software (version A.9.5.3.1). All arrays were of the same print run and hybridized on the same day. Microarray data were submitted to the National Center for Biotechnology Information's Gene Expression Omnibus (GSE17901).

Microarray quality control and data analysis.

Two arrays were excluded from further analysis for poor hybridization characteristics per feature extractor quality control metrics. The remaining array data were examined using Genespring 10 (Agilent Technologies). Hierarchical analysis clustering (HAC) and principal components analysis (PCA) were employed to confirm and identify appropriate treatment subgroups and eliminate outliers. Final analysis groups consisted of saline 7-day (6 arrays), 14-day (6 arrays), and 28-day (5 arrays); ANG II-treated 7-day (4 arrays), 14-day (5 arrays), and 28-day (6 arrays); and ANG II-treated CR 7-day (3 arrays).

Two-class, unpaired Significance Analysis of Microarrays (SAM v. 3.09) was performed in every combination with false discovery rate (FDR) < 1%(40). For gene list comparison, FDR < 5% cutoff lists were obtained as well, although all pathway, ontology, and network analysis made use of FDR <1% data only. Gene lists were probed for overabundant (P < 0.05, 2 genes/category minimum) pathways (Kyoto Encyclopedia of Genes and Genomes, KEGG) and Gene Ontology (GO) Biological Process groups using DAVID (14) against the Agilent MouseV2 (current mouse whole genome annotation) background.

Literature-based association networks were created to identify highly connected nexus genes by text mining, employing the 7-day treatment groups and the Agilent literature search plug-in (v. 2.69) for Cytoscape (v. 2.6.1), as previously described (1, 34). The term “nexus genes” emphasizes their central role in biological networks and distinguishes them from hub genes, in which connections are derived from network analysis centered around gene expression correlation.

An association network is derived from text mining of Medline abstracts and association identified between any two genes if they appear in the same sentence as an interaction verb as defined by the user context file. A series of subnetworks (independent of the experimental data) is then generated, and the expression values and significance of genes in our analysis overlaid visually and mathematically onto these networks. Context “OR” search terms included “aorta,” “aortic,” “atherosclerosis,” and “aneurysm.” Inclusion as a nexus gene required an arbitrary minimum of five neighbors with inclusion in at least five literature sources. Networks were individually curated to eliminate false calls based on alias mismatches. After overlaying expression values and SAM (d)-scores, we identified highly interconnected nexus genes. These were ranked either by mean significance (d)-score value for all subnetwork members or by a combination score derived by averaging the mean (d)-peripheral score with the nexus (d)-score. Selected differentially regulated nexus genes were examined by Taqman (Applied Biosystems) qRT-PCR to confirm observed array results using standard methodology, with normalization to Gapdh expression.

Immunohistochemistry.

Using frozen sections of mouse suprarenal aortic aneurysms obtained during the treatment time course, we performed immunohistochemistry (IHC) to verify the gene expression results for two nexus genes, Spp1 (osteopontin) and Adam17/TACE (TNF-alpha converting enzyme). Rabbit anti-mouse osteopontin (1:100; O-17, 18621 IBL-America, Minneapolis, MN) and anti-mouse Adam17 (1:200, ab2051; Abcam, Cambridge, MA) were incubated with tissue sections per standard protocol and visualized with biotinylated goat anti-rabbit secondary using a Vectastain ABC system (Vector Laboratories, Burlingame, CA).

Statistical analysis.

Nonarray data are expressed as means ± SE. Student's unpaired t-test and Tukey-Kramer ANOVA testing were used, with statistical significance at P < 0.05.

RESULTS

Treatment time-course.

Baseline aortic sizes were similar for all cohorts. ANG II treatment increased suprarenal aortic diameter throughout the 28-day course (Fig. 1), with significant differences from saline-treated at each time point monitored (P < 0.01) and from each preceding time point (P < 0.01) except for the period spanning 14 days to 28 days. After 7 days, the ANG II-CR group had the largest average aortic diameter (1.97 ± 0.21 mm) (P < 0.001). No significant size differences at any given point were observed within the non-CR ANG II-treated cohorts.

Fig. 1.

Maximal suprarenal abdominal aortic diameters over time course by treatment group. *Treated lumen size > saline treated (P < 0.01). Saline, saline treated (7 day, n = 18; 14 day, n = 12; 28 day, n = 6); angiotensin II (ANG II), ANG II treated (7 day, n = 18; 14 day, n = 12, 28 day, n = 6); ANG II CR, angiotensin II treated, with contained rupture identified at 7 days (n = 3). Error bars: ± SE.

Representative longitudinal and transverse ultrasound views and hematoxylin-eosin stains of the saline, ANG II, and ANG II-CR groups are depicted in Fig. 2. In the ANG II group, ultrasound and histology show a dilated aortic lumen, but relative elastin preservation and minimal thrombus visualization. In contrast, the ANG II-CR group displayed more prominent aortic dilation with segmental mural disruption. In the example shown, intramural thrombus appears as a hypoechoic focus. Examples of gross tissue appearance are shown in Fig. 3. Measurements were confirmed ex vivo after harvest. By histology, CR samples showed elastin degradation, medial rupture, and saccular accumulation of hematoma contained by adventitia.

Fig. 2.

Sample aortas observed by ultrasonography, and histology (hematoxylin-eosin stain). Representative images from mice infused with saline or ANG II for 28 days, suprarenal region shown. A, D, G: ultrasound longitudinal view. B, E, H: ultrasound transverse view. C, F, I: histological cross section. A, B, C: saline treated. D, E, F: ANG II-treated, abdominal aortic aneurysm (AAA) development. G, H, I: ANG II treated with aneurysm and CR, hematoma seen on ultrasound (white arrows), and histology (*). Note that I represents a section through the hematoma seen in G and H. L, lumen; blue line, maximal aortic diameter; white bar, 1 mm; black bar, 300 μm.

Fig. 3.

Aortic tissue collection and gross pathology. Aortas were harvested under dissection microscopy, and surrounding connective tissues were removed as much as possible. Presence or absence of any significant CR was examined by transverse incision of the suprarenal aorta before the tissue was flash-frozen in liquid nitrogen. A: saline-infused aorta. B: ANG II-induced AAA in the absence of CR. C: ANG II-induced AAA displaying CR. D: transverse section of C. Single arrows, left renal artery; Double arrows, hematoma. Scale: 1 mm.

Microarray expression profiling.

Strong differential expression was found between ANG II- and saline-treated aortas at 7 days. Identified were 6,192 unique genes at FDR <1%, with 3,471 increased with ANG II and 2,721 decreased, constituting 22.9% of the total unique genes on the array (for FDR <5%: 5,303 and 4,864 unique genes respectively). In contrast, 14-day time-point comparisons between ANG II and saline demonstrated dramatically fewer differences (52 total unique genes at FDR <1%, 283 genes at FDR <5%), and by 28 days very few changes were seen (7 genes at FDR <1%, 63 genes at FDR <5%). For ranked examples, see online Supplemental Tables.¹

Saline-treated samples showed consistent gene expression, with no differential genes identified between any time-point pair, clear intermixing between cohorts by HAC, and minimal scatter by PCA. PCA also supported the conclusion that the majority of the effects of ANG II on aortic gene expression in this model are short-lived. Furthermore, additional statistical analysis examining the relative contributions of data distribution, interarray variance, and effect-size changes on the presence of fewer significant genes over time showed that decreased treatment effect size was the predominant factor, a direct result of expression values in ANG II samples approaching those in saline samples. Significant differences in intrasample variance between ANG II- and saline-treated were not observed.

Ontologic/pathway analysis.

We performed ontologic pathway enrichment analysis of the significant 7-day gene lists, with particular attention to KEGG pathways and GO Biological Processes. This revealed widespread upregulation with ANG II treatment (vs. saline) of several broad categories, including those related to immune responses, matrix remodeling/cellular dedifferentiation, and cell signaling (Fig. 4). Specific upregulated pathways included cell cycling, apoptosis, Toll-like receptor (TLR), MAPK and p53 signaling, angiogenesis, fibroblast proliferation, and extracellular matrix (ECM)-receptor interactions. Many elements of immune system activation were increased, including genes related to T-cell, B-cell, and natural killer responses. This was accompanied by increased expression of numerous cytokines.

Fig. 4.

Pathway enrichment analysis at 7 days, ANG II vs. control. Top: Selected enriched KEGG pathways upregulated with ANG II treatment (vs. saline), ordered top to bottom from lowest to highest P value (EASE P < 0.05). Values shown are total unique upregulated genes per pathway over total number of genes per pathway on the array. Bottom: selected enriched Gene Ontology (GO) Biological Process categories for the same comparison. (EASE cutoff P < 0.05.)

The primary significant ontologic category representing genes downregulated by ANG II was “metabolism,” with 34 significant metabolism-related GO terms. Examples included the TCA cycle, glycolysis, and amino/fatty acid metabolism (not pictured). Lists of enriched KEGG categories and their genes are in the Supplemental Tables.

The 14-day samples showed ongoing upregulation with ANG II treatment for some of the same KEGG pathways, including ECM-receptor interaction, focal adhesion, and Wnt signaling. Notably absent from the significant ontology results were inflammatory/immune related pathways. The most significant GO terms were those related to ossification and vascular development, followed by ECM reorganization and regulation of cell-substrate adhesion. In particular, the top-ranked 50 upregulated genes are visibly dominated by collagens and related ECM genes such as Col1a1, Col1a2, Col2a1, Col3a1, Col8a1, Col8a2, Col11a1, Col12a1, Col16a1, Fap, Fbn2, Fndc1, Fgf18, Cthrc1, and Fbln2.

CR vs. saline and non-CR aneurysms.

CR morphology within the 7-day aneurysms was associated with major gene expression changes distinguishing them from either control (6,239 total unique genes: 2,643 CR > saline, 3,596 saline > CR, FDR <1%) or ANG II-treated aortas that showed no CR (1,798 total unique genes: 202 CR > non-CR, 1,596 non-CR > CR, FDR <1%). CR aneurysms displayed highly significant decreases from non-CR aneurysms in many transforming growth factor (TGF)-β/bone morphogenetic protein (BMP) family signaling genes (Fig. 5). Also decreased were ErbB, MAPK, and p53 signaling pathways, accompanied by a drop in focal adhesion and ECM-receptor interaction genes (Fig. 6). TGF-β/BMP pathway and ErbB signaling in CR aneurysms actually decreased significantly below expression levels in saline-treated aortas.

Fig. 5.

Transforming growth factor (TGF)-β pathway genes altered with CR. TGF-β pathway genes from KEGG that were significantly decreased in 7-day ANG II-treated AAA with CR (compared with non-CR, ANG II-treated AAA), false discovery rate (FDR) < 1%.

Fig. 6.

KEGG pathway enrichment analysis at 7 days, ANG II alone > ANG II-treated + CR. Ordered top-to-bottom from lowest to highest P value. Values shown are total unique upregulated genes per pathway over total number of genes per pathway on the array. (EASE cutoff P < 0.05.)

Nexus genes characterize ANG II-induced AAA.

Employing text mining of Medline abstracts, we generated individual gene subnetworks and awarded each an overall significance score. Genes with the highest significance scores were both highly interconnected and had numerous highly regulated neighbors and were therefore designated nexus genes. The top 80 ranked nexus genes were examined in detail (Supplemental Tables). They included a number of genes associated with inflammation. Among the nexus points identified were several chemokines (Ccl2, Ccl3, Ccl5, Cx3cl1, and Cxcl16), chemokine receptors (Ccr2, Cxcl16), and interleukins/receptors (IL6, IL11, IL1b, IL1r1, IL2). Matrix metalloproteinase (Mmp)3 and Mmp14 were present, as were the MMP-inhibitors tissue inhibitor of metalloproteinase (Timp)1 and Timp2.

A number of growth factors/receptors appeared as highly ranked nexus genes, including Tgfb1, Pdgfrb, Hgf, Met, Egfr, Igf1, Igf2, Igfbp1, and Igfbp5. Hematologic factors (Plau, Plaur, Thbs1) and a number of lipid-related genes (Apob, Apoc3, Lrp1, Ldlr) were present. Finally, some intriguing subnetworks surrounded nexus points Myd88, Spp1/osteopontin, Adam17/TACE, Txnip, Lox, Alox5, and Alox5ap (examples in Fig. 7). Differential gene regulation and protein expression of several nexus genes with aneurysm development were confirmed by qRT-PCR (Tgfb1, Myd88, Lox, Adam17/TACE, Spp1, and Igfbp1), and IHC of suprarenal aortas (Spp1/osteopontin, and Adam17/TACE) (Fig. 8). Enrichment analysis of the top 80 nexus points identified pathways that mirrored many of those found for the overall list of genes upregulated by ANG II at day 7, adding credence to the network-based approach (Fig. 9).

Fig. 7.

Highly ranked nexus gene subnetworks. A Medline-based association network of gene regulation in ANG II-treated AAA at 7 days (vs. saline control) was constructed using Cytoscape. (d)-Scores were derived for all genes on the array. Nexus genes were identified and ranked by averaging expression (d)-scores of immediate subnetwork neighbor nodes with the primary nexus (d)-score. Top left: Adam17/TACE subnetwork. Top right: Spp1/osteopontin subnetwork. Bottom left: Myd88 subnetwork. Bottom right: Lox subnetwork. Nodal green/red saturation corresponds to level of down/upregulation respectively. Gray, not significantly different from control at FDR < 1%. Blue lines represent direct literature-identified connections. Line thickness is proportionate to number of references.

Fig. 8.

Enhanced aortic expression of osteopontin and ADAM17 in ApoE−/− mice infused with ANG II for 7 days. Top: suprarenal aortas of ApoE−/− mice infused with saline (A, B, and C) and ANG II (D, E, and F) were examined for osteopontin (B and E), ADAM17/TACE (C and F), by immunohistochemistry. Negative control (A and D): samples incubated without primary antibody. Black bars, 50 μm. Bottom: selected nexus gene expression by Taqman qRT-PCR at day 7 of ANG II treatment. Average fold-change vs. saline treated shown ± SE, y-axis in log(2)-scale. All bars significant at P < 0.05.

Fig. 9.

Enriched nexus KEGG pathways, ANG II > saline (7 days). The top 80 ranked nexus genes were examined for overabundant KEGG pathways (P < 0.05). Number of nexus genes found within each pathway is shown.

The combination rank score described above favored genes that themselves were highly regulated. We also used a different approach, averaging the (d)-scores of peripheral subnetwork members without weighting toward the nexus. This yielded a few highly interconnected candidates that, while not differentially regulated themselves, had strongly regulated subnetwork members, possibly suggesting posttranscriptional processes. Among the highest-ranked nodes of this variety were Tweak (Tnfsf12), a cytokine known to induce apoptosis, and the endothelin receptors Ednra and Ednrb.

DISCUSSION

There have been many attempts to model human AAA development. As with rodent ATH models, there are obvious caveats in modeling this complex disease, especially in terms of rate of progression. The ANG II-treated ApoE−/− mouse is a commonly used rodent model that has been well studied because of many features it shares with human AAA, but until recently the model had never been subjected to whole genome expression profiling.

The aneurysm progression we observed was similar to that seen in other studies, although we specifically divided our dataset into AAA either with or without clear mural disruption, hematoma, and CR on ultrasound. Our mice were relatively young, which led both to a somewhat lower mortality rate and fewer CR events than are sometimes reported. Another study examining this model with ultrasound showed a more rapid early growth phase (3–7 days), but also combined their CR and non-CR values (2).

While the number of genes separating AAA from control at 7 days was quite large, the quantity dropped dramatically with an extended treatment course, despite progressive aortic dilatation and blood pressure readings at 3-day intervals that confirmed ongoing ANG II drug effect with systolic elevation of ∼40 mmHg, the latter also being a well-documented effect in the literature (8, 26). Although 6,192 unique genes distinguished ANG II-treated aortas at 7 days, by 14 days this had decreased to 52, and by 28 days down to only 7 at FDR <1%, nearly indistinguishable. Genes showing expression changes at 14 days were essentially a subset of those seen at 7 days, with 96.2% overlap. One key difference was the increase percentage-wise of upregulated collagen and ECM-related genes in the day 14 samples (compared with day 7), possibly reflecting a healing/scarring response and suggesting active fibrosis. Some of the decrease was likely due to varying gene expression response and intermouse variability at the later time points; however, analysis indicates that the phenomenon observed is largely related to decreased effect of ANG II on gene expression. As expected, saline-treated aortic samples were not detectably different from each other. Studies that rely only on samples harvested at a 4-wk treatment time-point may substantially underestimate expression changes involved with AAA development in this model, at least in younger mice.

Many of our ontology and pathway analysis results from the 7-day time-point generally concur with specific findings in both human AAA and studies using this rodent model (4, 23, 25, 39). Typically, focal acute macrophage infiltration of the aortic media begins within 48 h, persisting at least 10 days. By 4 wk, there is evolution toward a chronic lymphocytic phase, with fibrosis and vascular remodeling (33). The expression changes we observed in inflammatory pathways likely reflect this. Early wide-ranging and robust immune system activation was observed, including T-cell, B-cell, natural killer, cytokine/chemokine, leukocyte activation and transendothelial migration, and TNF signaling. ANG II treatment also induced other processes believed to be associated with AAA development, including angiogenesis signaling, fibroblast proliferation, and matrix remodeling, with attendant changes in cell-cell and cell-matrix adhesion genes (26). By 14 days of ANG II treatment, significant immune-related pathway changes no longer appeared, although genes from adhesion, matrix remodeling, and Wnt cell signaling pathways remained significantly increased. The question arises whether the temporal gene expression changes we observed represent alteration in local cellular makeup, as opposed to changes in gene expression from already present cells. Given the pathways involved it seems very likely that the expression signature derives from both infiltrating immune cells and dedifferentiating mural SMCs and that this accurately represents the pathobiology of the model.

The effects of ANG II on local renin-angiotensin signaling (RAS) pathways in aneurysmal disease have been explored by the developers of the mouse model, Lu and colleagues (review, Ref. 24). In our study, as would be expected, the model does show regulation of RAS-related genes. It is well established that treatment of vascular smooth muscle cells with ANG II downregulates expression of its own ANG II type 1 receptor (22, 28). Consistent with this, we found that at the 7-day ANG II time-point the Agtr1a gene was dramatically downregulated in aneurysmal aorta (8.3-fold compared with saline). A few other genes in the RAS pathway (KEGG ID: map04614) were also upregulated in the vessel wall including mast cell chymase 1 (Cma1) and cathepsin A (Ctsa).

Recently Rush et al. (32) published a whole genome analysis of ANG II-treated ApoE−/− AAA. Using whole aortic tissue obtained from 17 wk old mice after 4 wk of treatment with either saline or ANG II (1,440 ng·kg⁻¹·min⁻¹), they identified 1,030 unique transcripts showing differential expression, a considerably smaller set than we found at 7 days (possibly attributable to our observed return to homeostasis) but larger than our 28-day comparisons. Differences may be partly due to their use of the whole aorta rather than the suprarenal at-risk/diseased segment, their higher dose of ANG II, our decision to harvest the CR cohort at 7 days, or differing statistical analysis. In addition, our study focused on the mural/intimal regions of the aortic wall, as we attempted to remove all adventitial tissue from our array-bound aortas. Despite these efforts, some residual adventitial signal might have contributed to our results.

Notably, subjecting our 28-day samples to asymptotic t-testing at P < 0.05 with a 2.0-fold cutoff (as in Rush et al.'s analysis) rather than the more stringent permutative approach yielded a similar number of genes (787 unique gene symbols) significantly enriched for the same KEGG categories they identified. Of the differentially regulated genes Rush et al. found between ANG II-induced AAA and saline, 585 overlapped with our 7-day FDR <1% gene list (388 up in AAA, 197 down). Moreover, our pathway enrichment at 7 days also identified similar processes including focal adhesion, TLR signaling, leukocyte recruitment and activation, transendothelial migration, natural killer, and B lymphocyte receptor signaling, indicating earlier modification of these pathways than previously appreciated.

In contrast to previous studies we separately analyzed CR samples. Thus, beyond the additional processes identified by our time-course analysis, we identified pathways that provide insight into this clinically relevant phenomenon. While the CR cohort may represent acceleration rather than divergence from standard AAA progression, it was distinct in morphology and gene expression from the other cohorts.

The overabundance of TLR signaling with AAA development in the current study is intriguing. In large arteries, these receptors are primarily expressed on dendritic cells, activating immune responses through T-cell autoreactivity and cytokine production (11, 30, 35). TLRs may also participate in the proinflammatory effects of ANG II (16) and contribute to atherosclerotic plaque formation and vascular remodeling (6).

We identified upregulated expression for a number of TLRs with ANG II treatment at 7 days (vs. saline), including Tlr1, Tlr2, Tlr6, and Tlr7. Interestingly, Tlr4 was not differentially expressed, although it has been reported to be upregulated with ANG II in other models(42). Tlr3 showed decreased expression in CR-AAA (vs. both non-CR/ANG-II-treated and saline-treated), while Tlr1, 2, 6, and 7 were up in CR vs. saline, but not vs. non-CR/ANG-II-treated.

TLR signaling pathways are separated into two groups: a MyD88-dependent pathway leading to the production of proinflammatory cytokines with rapid activation of NF-κB and MAPK, and a slower TRIF-dependent pathway associated with interferon activation (18). MyD88-dependent inflammation is believed to be required for flow-mediated inward remodeling of conduit vessels (37), and its deficiency reduces ATH in ApoE−/− mice (27). We identified Myd88 as a highly interconnected and high-ranked nexus gene, providing further evidence for the involvement of MyD88-mediated TLR signaling in AAA development (Fig. 7).

Toll receptor Tlr3 may play a role in the etiology of AAA rupture. Tlr3 primarily operates via the TRIF-dependent pathway, which is opposed by TRAF1. Consistent with the observed expression described above for Tlr3, Traf1 was increased in CR aneurysms compared with saline control, while downstream pathway factors Tbk1 and Traf3ip1 were decreased. Tlr3 is expressed in aorta and carotids, but not temporal or iliac vessels (30). Furthermore, Tlr3 may be associated with TGF-β release, myofibroblast conversion, and stimulation of extracellular matrix production (35), processes that might stabilize a growing aneurysm, preventing rupture. As seen in Figs. 5 and 6, CR was accompanied by broad decreases in both matrix-related and TGF-β/BMP pathway genes.

A caveat in pathway analysis is that even though many genes in a given pathway are downregulated it does not necessarily mean the downstream genes need be. Both crucial pathway activating genes (Smad4, SarA/Zfyve9, Thbs1/thrombospondin) and inhibiting genes (Dcn/decorin, Ltbp1, Smad6, Smad7, Smurf2) were downregulated with CR. However, in this case downstream elements of the pathway that are targets of either Smad signaling or noncanonical pathway signaling were also downregulated, such as Rock1, Rock2, E2f5, Mapk1, Ppp2ca, and Ppp2cb, implying a overall comparative decrease in pathway activity despite the elevated levels of TGF-β. In fact, 30 genes from the downstream MAPK-signaling KEGG pathway were significantly lower in ANG II-CR than in ANG II without CR.

While CR is a somewhat rare event, an ancillary study was performed to examine TGF-β signaling in CR aortic segments using the same protocol and identified two CR mice at day 14 of ANG II treatment. With Smad2 and Smad3 protein levels as controls, Western blotting showed decreases of 46 and 70% in phospho-Smad2 and phospho-Smad3, respectively, when compared with non-CR aneurysms at the same time point, suggesting that CR may be associated with decreased canonical TGF-β signaling (data not shown).

Note that also widespread decreases in BMP-pathway genes (part of the TGF-β superfamily) occurred, including all three of the Bmp receptor genes, the activating ligands Bmp4 and Bmp6, and the downstream Id4, suggesting global downregulation of the BMP system with CR. As the BMP pathway is associated with vascular calcification and ossification, loss of this pathway might deprive the aorta of wall stabilization, leading to higher risk of rupture.

We identified a number of nexus genes within the expression network. This set of genes contained many with well-known associations with AAA, including numerous proteases, cytokines, interleukins, and their receptors. These findings not only serve as verification of both the method and the centrality of the inflammatory process to AAA development but also prioritize specific genes that may play important regulatory roles.

Nexus genes may help to address the question: to what extent is AAA an atherosclerotic process, as opposed to a more cell cycling/ECM remodeling process? In addition to their role in modeling AAA, ApoE−/− mice are also used to model ATH. Some of the aortic gene expression changes observed with ANG II treatment likely represent rapid progression of ATH. We compared our day 7 ANG II treatment gene lists and GO categories with the ATH signature derived by Tabibiazar et al. (36). As expected several of the mouse disease classifier genes they derived overlapped both with the ANG II-upregulated list and with some of our highest-ranked network-derived nexus genes, including Spp1, Timp1, and Ccr2. Examples of overlapping enriched GO categories included: “cell death,” “cell adhesion,” “chemotaxis,” “immune response,” “inflammatory response,” “ossification,” “locomotory behavior,” and “negative regulation of angiogenesis.” Some nexus genes associated with human coronary ATH severity (20) were also differentially regulated in our study, such as Ccnb1, Bclaf1, and Abcc2 (up with ANG II and with increased severity), and Ryr2 (decreased with ANG II and associated with milder disease).

In some ways, the gene signature of ANG II-induced AAA also resembles that of injury-induced vascular remodeling. Our networking methodology was previously employed to compare gene expression between human in-stent restenosis (ISR) and de novo ATH (1). In that study, the predominant gene signature of ATH was inflammatory/immune-related, while ISR primarily involved cell growth and ECM genes. Intriguingly, every differentially regulated nexus gene that characterized ISR (vs. ATH) was upregulated at day 7 in ANG II-treated aortas (vs. saline) (Col1a1, Col1a2, Col3a1, Col5a2, Map4k4, Gli3). However, of the six, only Col1a1 also constituted a AAA nexus.

In contrast, an ATH-defining nexus was also identified: Adam17/TACE, the TNF-α-converting enzyme. This molecule is thought to regulate inflammation by releasing cell surface transmembrane proteins, such as L-selectin, TNF-α, and its receptors TNFR1 and TNFR2. Adam17 was upregulated in ATH (vs. ISR) along with many of its subnetwork members, and the same was true in ANG II-treated aortas. Adam17 IHC confirmed increased expression at 7 days within aneurysmal aortic media. As a prominent nexus, Adam17 may represent an attractive therapeutic target for stabilization of evolving AAA, with the goal of modifying the manner in which the vessel remodels.

The most prominent known substrate of Adam17 is TNF-α. Plentiful evidence in the literature ties TNF-α to abdominal aneurysm development, in humans (where levels are increased locally) as well as in mouse models. Abrogation of TNF-α through inhibitory antibody therapy decreases aneurysm development in the murine CaCl₂ treatment AAA model (43). One study, aimed at evaluating the effects of TGF-β, examined TNF-α in the murine ANG II AAA model (41). While infliximab inhibition of TNF-α did not significantly affect AAA development in the context of anti-TGF-β treatment, a clear trend toward benefit was seen, despite their use of C57BL/6 mice rather than the more inflammation-prone ApoE−/− line used in our study.

Another potential use of nexus genes would be to provide informative biomarkers. One gene identified as a prominent nexus, Spp1/osteopontin, was found to predict AAA presence and growth in a limited patient cohort using serum levels after adjustment for other risk factors (10). Furthermore, osteopontin-deficient mice have attenuated AAA and ATH development in the ANG II model (3). We add to previous reports by showing increased aortic osteopontin protein and gene expression with aneurysm formation as early as 7 days into the treatment course.

A novel finding was the identification of Lox/lysyl oxidase as a highly ranked upregulated nexus with no previously described role in AAA. Lox protein is known to be secreted extracellularly, suggesting utility as a biomarker. Lox may play a key role in vascular homeostasis, as it is chemotactic for SMCs and monocytes, maintains ECM stability, and participates in vascular remodeling (31). Lox showed significant downregulation with CR, as did Loxl1 and Loxl4 (Lox-like). Loxl1 single nucleotide polymorphisms are associated with spontaneous cervical artery dissection, while Loxl4 has not yet been connected to known aneurysm pathology but is a target of TGF-β signaling (19, 21).

While our clearest findings involve genes that were upregulated in AAA at 7 days, an equally large number of transcripts were downregulated. The obvious pathway signals involve aspects of energy metabolism, including oxidative phosphorylation, TCA cycle, fatty acid metabolism, glycolysis, etc. (Supplemental Tables). Peroxisome proliferator-activated receptor (PPAR) signaling was downregulated as well, mirroring the findings of a 2002 gene study utilizing the ANG II/ApoE−/− model (39). Moreover, Jones et al. (17) have recently demonstrated that the PPAR-γ agonist rosiglitazone reduces development and rupture of AAA in the same model.

Also prominent among the downregulated genes were alcohol and aldehyde dehydrogenase enzymes (Adh1, Adh4, Adh5, Aldh1a7, Aldh2, Aldh4a1, Aldh5a1, Aldh6a1, Aldh7a1, Aldh9a1). Human blood vessels contain large amounts of these proteins, and class I ADH activity has been found to be lower in human AAA than in healthy aorta (15).

We did not attempt to independently profile individual cell types within the aortic wall, and our data therefore represent a composite portrait of expression, with the relative contributions of inflammatory cells, endothelial cells, and smooth muscle cells remaining unknown. While there are obvious differences between human AAA development and that observed in the murine ANG II model, many parallels also exist including similar histopathology and similar specific gene/protein expression (4, 5, 33). Our study would suggest that the early phases of murine AAA development in this model will be the most fruitful in modeling human disease. In addition to our results suggesting potential pathways and genes/gene complexes to be targeted therapeutically, many of the nexus genes identified might also be secreted and serve as circulatory biomarkers for detection of AAA and disease progression.

Our study was limited by several mice expiring due to rupture and being unable to undergo expression profiling. However, the few mice that survived with CR constitute a particularly interesting group. In humans, AAA rupture represents a significance source of mortality, and, apart from aneurysm size and growth rate, no clear clinical predictors exist to stratify rupture risk. The genes/pathways that differentiated nonruptured AAA from CR could suggest ways to stabilize human AAA or to detect impending rupture.

In summary, we have performed detailed expression profiling during the time-course of AAA development in the ANG II-treated ApoE−/− mouse, confirming and expanding upon previous work. We observed wide-ranging increases in inflammatory genes with aneurysm formation, as well as numerous other processes. Subanalysis of CR aneurysms yielded key cell signaling pathways that change in association with this phenotype. Our experiments showed profound gene expression changes early in the treatment course, which decreased considerably over subsequent weeks, suggesting that aortic tissue adapts to prolonged ANG II infusion. Furthermore, network analysis discovered subnetworks of highly connected genes that may constitute biomarkers or therapeutic targets.

GRANTS

This study was supported in part by National Institutes of Health Grants 5K08 HL-080567-03 (J. M. Spin) and 1P50HL-083800-05 (R. L. Dalman overall PI, PST-PI Project 2).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).